Open Source--The Unauthorized White Papers Chapter 10, The Platform: Software and Hardware

Open Source: The Unauthorized White Papers

Chapter 10

The Platform: Software and Hardware

If Open Source software were available only from Internet downloads, and installable only by experts, there would not be much of it around. From the beginning there were services that saved the time and difficulties of downloading and installing the early FreeBSD and Linux by providing coordinated, tested sets of files that worked together. From this beginning, hardware services sprang up that supplied equipment with the systems ready-installed, and paid support systems grew up as well. To these vendors, add custom software developers and systems integrators, and a whole industry is growing up to support Open Source software and users.

This chapter takes a look at software and hardware and service businesses that have been built on Open Source software. Because the market is so new there are opportunities to spot needs that are not being filled, whether in products or services. This chapter provides some of the categories that have already emerged.

Linux Distributions

Walnut Creek CDROM (http://www.cdrom.com) has been instrumental in getting Open Source distribution going. The company published the earliest widely distributed Linux CD (Yggdrasil) in 1992, and has been distributing Slackware Linux since that same year. In 1993, Walnut Creek began publishing FreeBSD on CD-ROM; recently the FreeBSD organization (http://www.freebsd.org) has begun publishing their own CD-ROM at the FreeBSD Mall (http://www.freebsdmall.com), as have other publishers in the U.S. and abroad. Walnut Creek recently merged with Berkeley Software Design (BSDI), who supply the proprietary BSD OS (http://www.BSDI.COM/).

Distribution, Publisher

A Linux "distribution" or "distro" consists of the Linux kernel and several hundred additional files that together form the Linux operating system or platform. Collecting these files from the various Open Source projects that originate them and make them available on the Internet, tuning and testing all the components so that they work reliably together, and providing a means of installation is the work of projects like Debian and for-profit companies like Red Hat, SuSE, Caldera, TurboLinux, MandrakeSoft, and others.

Walnut Creek CDROM does not prepare the Debian distribution (called Debian GNU/Linux), but does publish it. Anyone is free to take that or any other Linux distribution and publish it as well. This leads to the curious economic model explained in the following pages.

Besides earning money on distributing CDs of software in which it has no property rights, Walnut Creek maintains one of the largest free software download sites in the world (ftp://ftp.cdrom.com), where many of the same software distributions, ready for use, may be had free for the downloading. The solo server handles 750,000 visitors a day (6,000 may visit at once) and demonstrates the strength of the FreeBSD operating system and the Open Source economics by which free software distribution enlarges the market — where, in turn, an expanding number of CDs are sold. Visitors download nearly a terabyte of software daily over the constantly full pipeline; a one-day test of the maximum demand showed it to be 1.39 terabytes, which were then served.

Major Linux distributions such as Red Hat and SuSE show a similar pull on their sites for free downloads of the software that they package and sell on CDs; in addition, other publishers distribute unofficial versions of the major distributions. Neither free downloads nor republished versions make a penny for the originators. At an earlier stage of its growth, it was worth Red Hat’s while to cooperate with the publisher Macmillan in issuing an Official Red Hat Linux (as distinguished from an unofficial copy): Macmillan pushed Red Hat into sales channels that it could not otherwise reach. After an initial push for market share, however, it no longer made sense for Red Hat to see Macmillan earning money on boxes of software that Red Hat now had the power to push into the same channels and sell. As a result, the agreement lapsed, and Macmillan picked up the MandrakeSoft Linux distribution in its place.

All the major distributors see organizations like Cheap Bytes (http://www.cheapbytes.com) and the Linux Mall (http://www.linuxmall.com) and others selling $2 or $3 CDs containing unsupported versions of their software. In the cases of Cheap Bytes and Linux Mall, these are retailers who sell the official versions as well, along with other products such as CD-ROM collections of documentation, but it is possible for any entrepreneur to publish the major distributions for a couple of dollars and figure out another way to make money out of the opportunity, perhaps by selling mailing lists of the customers. If you have followed this model so far, you are ready for the next surprise: some of the major Linux distributions set up mirror sites or send master CDs to the other publishers to encourage them to continue the practice. The rationale is that the cheap CDs provide a way for users to try out the distributions; when the user is ready to take one seriously, the user will naturally want to buy the full distribution, which will include some sort of support and some proprietary applications, such as an office suite that might have to be paid for anyway. Thus the cheap CDs serve to distribute sample copies of the various distributions at no cost to their originators and at some profit to the publishers.

It is possible to take an existing Linux distribution and piggyback on it to create value in a new distribution. MandrakeSoft (http://www.linux-mandrake.com) began by selling a distribution that added value to the Red Hat 5.1 distribution by adding the KDE desktop. MandrakeSoft has since evolved its own distribution called Linux-Mandrake, tuned especially for the Pentium chip, but the company is keeping the distro compatible with Red Hat, whom it sees as the market leader. In turn, Mandrake has seen its distribution repackaged by various vendors. For instance, NECTEC (http://www.nectec.or.th/home/index.html), a Thai government-sponsored project to improve information technology in Thailand, together with the Thai Linux Working Group (http://linux.thai.net/MaTEL/), has developed a Thai version of Mandrake originally called MaTEL, or Mandrake & Thai Extension-Linux, and since renamed to Linux-TLE. China has its own simplified-character Chinese version, Red Flag Linux, which runs on the Compaq Alpha and Intel chips. Although all this fragmentation makes it appear that Linux will die the death of a thousand distros, the very license that makes it possible (see Chapter 6, "The GNU GPL and the Open Source Initiative") also exerts a centripetal force that tends to iron out the differences among distributions (see Chapter 7, "Complications of Open Source Licensing"). The very improvements that Mandrake adds to the Linux mix can be taken freely by Red Hat for its next distribution, and the same cross-pollination process holds for the other distributions.

Beyond technical distinctions, which may simply be appropriated by competing distributions, marketing prowess determines the viability of distributions. Marketing promotions may also be copied, of course. Red Hat originally positioned itself as cutting-edge Linux technology for developers; as it pushed more into the mainstream it was unwilling to abandon this part of the market, and set up the Raw Hide program to make available on-line the latest technology that was being developed in the Red Hat shop. MandrakeSoft has a similar on-line program, called Cooker. And as a piggyback on the Raw Hide program, there is a Japanese distribution called Kondara MNU/Linux (http://kondara.sdri.co.jp), based directly on Raw Hide.

Specialization in a niche market can protect a distribution from its larger competitors. Kevin Fenzi’s Red Hat Über Distribution (KRUD) — its motto is "We take Red Hat and turn it into KRUD" — collects the latest application errata, system patches, and some useful material not in the Red Hat distribution and builds an updated version of Red Hat Linux every month (http://www.tummy.com/krud/). Fenzi is a security expert, and the chief motive for purchasing KRUD is a desire to make a system secure by applying the latest patches. The practice of monthly rebuilds and CD distribution of them, along with the modest price ($5 for a single month’s issue, or $55/year) ensure that Fenzi will be able to keep this business for himself until someone comes along who can do a better job cheaper, and convince the buyer of this.

As a result of its level playing field, Linux finds itself in the position of a commodity; how could it be otherwise, given its stated goal of giving the most benefit possible to the user (consumer, customer) at the least cost? The biting irony of the situation is that Open Source fans, who tend to disparage Microsoft success as a result of superior marketing, and not of superior technology, find the success of Linux, and particularly of any one distribution of Linux, heavily dependent upon marketing. Accordingly, the Linux distributions have been plying the tools of marketing. They are targeting various market segments (and most are aiming at the largest and most lucrative, the ease-of-use segment), doing their best to understand and respond to the market needs, and adding ancillary products (support, training, and such) to make their distributions into a complete product. Services are traditionally one way to differentiate a commodity product from its competitor.

The strongest differentiator, however, is branding, and Red Hat’s success among the Linux distributions has been the result of Bob Young’s early and deep understanding of marketing and branding. Before there was a Red Hat, he watched the sales of Linux distribution CDs in his own business, and noticed that Patrick Volkerding’s Slackware distribution sold much better, and at a higher price, than its competitors. To techies! Here was objective (and bankable) proof of the power of branding. The current Red Hat company was born when the man who understood the market found the man with the software to market. Young enjoys using the example of Heinz ketchup as the example of the power of branding, and has used the example so often that the current revival in Heinz sales (after a brief eclipse by salsa) might possibly have felt some effect from Young’s speeches.

Branding power comes with its own built-in limitations. Knowing when brand extension fades into brand dilution is the mark of a superior product manager. A hardware manufacturer, for instance, wanting to install a branded Linux distribution on his machines might want certain changes made that would result in the brand-owner’s refusal to extend the brand name to cover such a version. And Corel Corporation, in the search for a Linux distribution to bundle with its own software application products, eventually settled on launching its own Corel Linux OS distribution (based on Debian).

Custom Software

The business model we have seen previously is that the originators of distributions take publicly available code and package it in some manner for further distribution. Some companies are able to follow this model with software other than the operating system, as the two examples that follow show.

Cygnus Solutions

Cygnus Solutions (http://www.cygnus.com) provides many developers with employment doing what they love: writing Open Source code. Thanks to the recent sale of Cygnus to Red Hat, some of these developers have found that they have become wealthy doing what they love.

Cygnus is the home of some of the most important development tools, both for the Open Source community and for many proprietary vendors. Michael Tiemann and two others founded the business to provide support and customization for the GNU C Compiler (commonly called GCC, or more often, gcc, in the lowercase UNIX manner), and for the GNU C libraries associated with it. Over the years the developers at Cygnus have ported these tools (and others) across many platforms. Most of this work has been paid for by vendors who wanted the tools ported to new languages or chips so that developers would be able to write software for them. All of this GNU-tool related code is delivered with source code under the GNU GPL, and the copyrights are automatically turned over to the Free Software Foundation.

Cygnus sells a state-of-the-art version of the tools called GNUPro, and also a formerly proprietary tool called Source-Navigator, made Open Source after the merger with Red Hat. A third product, Code Fusion, makes use of both the Open Source GNUPro and the closed-source Source-Navigator. Another tool, called Cygwin, is distributed under a license like the Troll Tech QPL: the product is Open Source if the user uses it to develop Open Source products; if the user uses it for proprietary products, a license must be purchased.

Custom projects (such as the ports mentioned above) will have their source code released as part of the GNU tools; this is generally after a delay. The time is used to thoroughly test the customizations on the customer’s site. In addition, the source code may be held back for a period until the customer is ready to release the product (such as a new chip), at which time the code is released.

Cygnus has become so successful as the first enterprise founded on developing GPL’d software, that by now it has written about 70 percent of the code in gcc and employs about 80 percent of the gcc developers. Coordinating the continued development of the GNU C tools was too complicated and demanding for the slender resources of the Free Software Foundation, which agreed to place management of the project in the hands of the gcc Steering Committee. Cygnus turns over copyrights for new code to the Free Software Foundation, and no more than 45 percent of Steering Committee members may be Cygnus employees.

Now that Cygnus is owned by Red Hat, the Steering Committee is looking at measures to make sure that Cygnus/Red Hat employees do not dominate it. Red Hat, which likewise turns over copyrights for GNU software improvements to the FSF, proudly states that it employs five of the eight major contributors to the Linux kernel (but only 5 percent of kernel developers work for Red Hat/Cygnus). By acquiring Cygnus, Red Hat gains a reach into non-Linux platforms, as well as the embedded expertise and tools of Cygnus, which has a real time embedded OS called eCos.

Precision Insight (PI)

Just as Cygnus Solutions found a niche with the GNU software development tools, Precision Insight (http://www.precisioninsight.com) has found a niche with drivers for hardware devices, particularly 3D graphics boards. Hardcore Linux developers are devoted to fast video games, and they resented the lack of interest by graphics card companies in writing drivers for Linux. When the firms did supply such drivers, they were never accompanied by source code because the companies sought to protect any clues to their 3D secrets. The Linux community often had to provide their own drivers through laborious reverse engineering.

Several forces worked to change this situation. The growing number of Linux users caught the hardware companies’ attention, and Red Hat invested money to provide some of the important laptop graphics cards with Linux drivers, knowing that the ability to run Linux on laptops would help grow the Linux market. Red Hat hired Precision Insight to write the drivers; the challenge was to produce the drivers along with source code for them under the Open Source X License. In some cases the Open Source drivers load a 2D driver for the graphics card; this driver is then capable of loading another driver that will handle 3D, and for which source code is not available. A longer-term solution is to persuade the hardware vendors to show enough of their technology so that an Open Source middle layer can be written for 3D applications to call. This Direct Rendering Infrastructure (DRI) will be incorporated into the Linux kernel, and will have the added benefit of directing the rendering to be done by the graphics board, rather than by the software.

Opportunity

Software drivers for new peripheral hardware have been a problem in the Open Source world for two reasons. Although the drivers themselves must be binaries to function, Open Source users want source code for the driver. The manufacturers, however, seek to protect their proprietary advantages by hiding their interfaces. There is a further problem for Linux because of its ambitions to cover both high-end and low-end markets. While a popular operating system (such as Windows 98) need only cover the lower end of the printer spectrum, and a high-end product such as Solaris needs only to cover the upper end, Linux must cover them all. There may be room in the market for a company to specialize in writing printer drivers, just as Precision Insight has specialized in writing graphics device drivers.

Not only does the steady stream of new hardware devices result in work for Precision Insight; the company also optimizes current Open Source drivers for software vendors who want high performance on their particular application. This work is added to the Open Source driver and improves it, since there is a very large set of paths through OpenGL that might be implemented in a driver, but which are unlikely to be implemented unless someone actually has a purpose for the path.

OpenGL itself is a high-performance graphics interface originating at SGI. Because licenses for it are so expensive, an Open Source project called Mesa implemented the interfaces. SGI realized that Mesa could help extend the use of OpenGL in a market that SGI could not serve without lowering prices and forgoing the high income from the upper end of the market. To keep the Open Source market from forking OpenGL, SGI helped fund development on Mesa. The intention is that programs written for the one will run efficiently on the other. OpenGL is not the only standard feeling uneasy in the presence of the popular and free Mesa. Every copy of Linux carries a free version of the X Window graphics software (a graphics server), XFree86. The Open Group, which keeps the standard for X Window, has to take into account the possibility that XFree86 could fork the X standard. As a consequence they have granted the XFree86 project an "honorary" membership, waiving the usual large fees for vendor membership. For both X Window and OpenGL there are many more free copies than proprietary copies of these important standards.

If it is a sign of the great momentum of Open Source projects that these concessions have been made for Mesa and XFree86, it is likewise a sign of the strength of Precision Insight that both the head of the Mesa project and the head of the XFree86 are Precision Insight employees. As in the case of Cygnus and Red Hat, those who believe that such projects can be influenced by employers should remember how easily top developers can find new positions. PI can’t dictate to its Open Source developers, but it can use its knowledge of the market, its financial resources, and their programming expertise to find what the market wants, and commit to deliver and support the products. By hiring the developers, PI gains influence over the scheduling of projects (by committing resources to ensure their completion). There is a similar dynamic at work in Red Hat’s hiring of so many Open Source developers, and in its acquisition of Cygnus. Commitment to delivery and support are necessary to win the corporate market.

sourceXchange and Cosource.com

Cygnus and Precision Insight are also examples of Open Source developers doing the work they love — writing Open Source code — but they do it as employees of corporations. There are many developers who like to work more independently, and there are programs starting up for them, too. The value of the Internet is not so much that it eliminates the middleman, but that the middleman becomes an organizer of information; the selection and organization of that information constitute the middleman’s value. In the case of sourceXchange and Cosource.com, these businesses organize the needs of software users and the resources of software developers and create a market in which the parties may meet directly. sourceXchange (http://www.sourcexchange.com) and Cosource.com (http://www.cosource.com; recently purchased by Applix) operate at the wholesale and retail ends, respectively, of the their business model, but, true to their Open Source origins, there is no barrier to their borrowing each other’s business practices and entering each other’s niches and competing directly.

Tim O’Reilly, having demonstrated that it was possible to build a successful business publishing books on Open Source software, began a venture called Collab.Net to take advantage of business opportunities in the Open Source world. Brian Behlendorf, a veteran of Apache development, has begun its first enterprise, sourceXchange. A client wishing to sponsor the writing of software at the site registers the specification for the software, the price offered, and the Open Source license that will be used for distribution of the finished software. sourceXchange posts the RFP (Request for Proposals), and helps the sponsor select a likely proposal by a developer (or developer group) from among the respondents. Developer and sponsor agree on the work and the payment, and for a percentage of the fee, sourceXchange provides a project manager or intermediary between the two; this person eventually passes the job as fit for the sponsor and authorizes the payment to the developer. At the end of the project, both developer and sponsor publicly post evaluations of how each found working with the other, and the software is issued under some sort of public license, free to all users. Although the projects are sponsor-driven in the sense that they are sponsor-funded, there is a place on the site for developers to suggest projects they have completed, are working on, or would otherwise like backing for. Under the sourceXchange model, project sponsors tend to be corporations that fund the entire project. Hewlett-Packard was an early sourceXchange sponsor, and as a consequence has already had a project completed.

In contrast, the Cosource.com approach emphasizes the community as project sponsor. While Cosource.com fills the same managerial and buffer role that sourceXchange provides between sponsor and developer, Cosource.com lets individuals post projects they would like to see coded, and the prices they would be willing to pay for them. These individual desires are aggregated at Cosource.com, and the responding proposals by developers are also tracked there. The sponsors of a project agree on a proposal, and Cosource.com provides a manager/intermediary between the two groups; like sourceXchange, Cosource.com takes a portion of the fee for management. When finished, the project is released under an agreed Open Source license.

A number of interesting economic facts emerges from these two companies. The project sizes, measured in money, are larger at sourceXchange, sponsored as they are by individual firms. At Cosource.com, the amounts are smaller. The sponsors of Cosource.com projects are smaller, of course, but there are two forces likely to drive prices down in this model. The first is that sponsors know that it is very likely that what they are seeking is out there somewhere, and that the developer of it needs only a little financial coaxing to clean it up and make it presentable for public appearance. The second, and more interesting reason, is that because this model for paid software development draws directly on international resources via the Internet, the pool includes not only highly-paid American programmers, but others from around the world. Russia, for example, contains highly capable programmers for whom small amounts of dollars are significant rewards.

One of the hopes of the Open Source community rests in the global nature of the movement. Linux is already a bigger success overseas than it is in America. Knowing the ability of young people to delve deeply into technology, the Open Source community is eagerly waiting for the Open Source programming power of South America (for example) to be felt. sourceXchange and Cosource.com will serve as funnels or lenses to concentrate that power as it emerges, and the economic power of the auction model of these sites will ensure its long-term viability. The only proviso is that the broker or middleman agents give real value for the share that they receive.

Hardware and Integrated Systems

Although Open Source is thought of as a software movement that uses ordinary Intel machines, it created a hardware market practically from the beginning. The evolving market is growing from specialized hardware (such as disk drives and communications boards) to specialized machines (heavy-duty servers) to general-use machines (laptops and desktops with Linux pre-installed). As Open Source takes root and then passes into wider use, the hardware market will grow along with it. The Open Source hardware market will be much like the hardware current market: vendors will turn increasingly to service to build a viable business.

Hardware

Long before such large companies as Compaq, Dell, Hewlett-Packard, IBM, SGI, and Sun offered Linux on their hardware, there were small companies that hitched their stars to Linux. One was Cyclades Corporation (http://www.cyclades.com), which began by offering multiuser serial port cards; Linux users who wanted to set up networks bought them, particularly small ISPs who attached a dial-in modem to each port. Cyclades then moved on into routers for Linux.

Such specialized hardware suited a specialized technical market, but a couple of years later, in 1996, Cosmos Engineering Company (http://www.CosmosEng.com) was the first to offer a hard drive with Linux installed on it, ready to put into a machine and run. Cosmos (and other companies) also offered desktop machines (workstations) and Web servers with Linux on them, ready-to-run. Eventually the Web server aspects of these machines were packaged, minus the peripherals (keyboards, monitors, and the like) as ready-to-use plug-in Web servers, such as the Qube Microserver from Cobalt Microserver (http://www.cobaltmicro.com).

The hardware market is notorious for its low margins, which are a result of its commodity status. If a vendor offers superior quality (high-quality matched parts as the foundation of each system), service (installation, maintenance, guarantees, and such), and expertise, he or she can claim higher margins than competing commodity items. Accordingly, the successful Linux box makers are not merely dropping an operating system onto a screwdriver-assembled box, but are taking advantage of the Open Source nature of Linux or BSD to tweak the system for optimum performance on the hardware. The vendor may also choose a particular distribution of Linux or BSD for its suitability for the intended uses of the hardware.

VA Linux Systems (formerly VA Research; http://www.valinux.com) made staggering first-day gains when its stock first came to market. Part of that value came not just from hardware or software expertise, but from the company’s having hired a string of luminaries from the Open Source world, some of them from Red Hat. VA Linux, like Red Hat, employs these experts to improve its products, and like Red Hat, it releases their work to the community under the GPL. VA Linux originally installed the Debian GNU/Linux distribution on its machines, but found that customer preference ran heavily to Red Hat. Although it has worked to promote the Debian distribution by providing commercial support and helping to push a packaged version of the distro into retail channels, VA Linux listens to its customers and consequently offers all its machines with an optimized Red Hat 6.2 installed.

Some of the other firms having as their main business the building of computer systems with Linux or other Open Source software installed are Penguin Computing (founded by a former VA Research employee, and not to be confused with Penguin Systems, a Linux consulting company; http://www.penguincomputing.com); ASL Workstations; and Atipa Linux Solutions (http://www.atipa.com). Laptops, however, continue to be something of a problem in the Open Source world.

The problems with laptops originate in their hardware, which is always being changed, even while the laptop model number remains the same. The Open Source user wants to be sure the hardware will have drivers that will support Linux. Red Hat had to invest some money to have Open Source drivers written for the NeoMagic graphics boards, which are used in many laptops. Another hardware obstacle is the WinModem. The hardware vendor omits a chip or two that does the usual work of the modem (this makes it fit more easily into a laptop, and lowers its price), and instead passes these functions to the operating system. Since the operating system chosen by the vendor is Windows, Linux is unable to make the hardware calls it expects of a modem, and cannot run the Windows software expected to stand in for the missing hardware. A Linux software workaround for this problem may arrive shortly, however — another example of the Open Source community coming up with what it needs.

Of all laptops, the Sony Vaio Z505 is currently high fashion among Open Source coders; it is light and portable, and provides no hardware obstacles to running Linux. Nevertheless, it does not come with Linux installed. IBM was the first laptop company to offer factory-installed Linux (on its laptops without WinModems, of course). A new company, Tuxtops (http://tuxtops.com/), has just been formed to meet the needs in this market.

Integrators

System integration/value-added reselling is presently a wide-open field in Linux. Competition is tight in established platform markets (such as Windows), particularly because hardware vendors are seeking to increase their margins and bind their customers more tightly to them by supplying integration services. Consequently, the traditional value-added resellers (VARs) and system integrators are under increasing pressure from the very firms that sell them equipment. Because the Linux market is so new, there are few players here, but this does not mean that the same pressures of vertical integration are absent.

VARs and Systems Integrators

The VAR and integrator categories are often used interchangeably because there has always been some overlap between them, and because distribution channels themselves are becoming more integrated. Traditionally, the integrator is a bidder on a new project, and because government agencies purchase by bids and proposals, and selling to the government is specialized work, the term integrator frequently refers to government contractors who supply everything on a project from software to cabling.

The term "Value Added Reseller" (VAR; sometimes still called "dealers") on the other hand, originates in retail sales; the customer who needs a modem orders it from a local VAR. Because every sale involves some level of service, and because the more service, the better the price and profits the reseller can command, the VAR adds value by suggesting the most appropriate equipment and software, and perhaps helping to install it. From there it is one more step to offering support, and then to supplying complete systems. So the VAR becomes a systems integrator. Although the term generally means someone who sells to private businesses, as opposed to government sales and service, there is a tendency now to refer to the really large firms as "Global Systems Integrators" (GSIs). GSIs are the very large worldwide firms such as Andersen Consulting that deal with both government and private clients.

Systems integrators tend to be larger-scale businesses, able to carry the overhead of government bidding and contracting, and employing a wider range of in-house experts. VARs, on the other hand, include many small operations, and may settle down in close relationships with a limited number of customers in a particular business domain or niche. The VAR’s particular strength is that the customer may not have much computing expertise, and will thus depend on the VAR for guidance.

The term consultant applies to anyone hired for help with a computer system; the consultant may or may not be a dealer or reseller or connected with one.

In the first place, several Linux hardware manufacturers, such as VA Linux, offer systems integration services. In addition, they offer support. They are probably well down the road to offering not only complete systems to customers, but also installation and financing as well, as some hardware vendors in the established markets already do. Not only are the hardware manufacturers getting into vertical integration, the distro vendors are doing it also. Red Hat, for instance, is helping to wire up the 3,000 terminals in Auto Zone stores with Linux; since the licenses can’t really be sold, this is a pure service venture. There are also partnerships between distros and hardware manufacturers, and the major distros are either selling 24-7 support or partnering with third parties to offer it in their name. But these combinations, emulating the DEC or IBM operations that came before them, are not necessarily a threat to the small integrator, because their focus is on large customers, and their services are priced accordingly.

The spread of Linux will draw new players into both the large-scale and the small-scale spheres of operations. The one critical question is Linux expertise. Until their customers call for it, the high-end operations such as Andersen Consulting cannot be expected to offer Linux services, but they do have the advantage that when the call finally comes, they have UNIX experts on board who will be able to switch easily into Linux. Smaller integrators, on the other hand, are not so likely to have the UNIX or Linux expertise. The potential small integrators on the other hand, that is, the Linux gurus itching to start small businesses, and the recent Linux-loving college graduates, rarely have the business domain expertise or the business experience to start an integration business. Nevertheless, if existing Windows VARs were to decide they could run more profitable and less harried businesses by switching part or all of their customers’ operations to Linux, they would be in an excellent position to profit from any coming Linux wave.

Becoming a Dealer or Consultant

Linux experts looking for help in breaking into the VAR/integrating business should look at the Consultants’ Resources at the Bynari Systems Web site (http://www.bynari.com). By offering this help, Bynari seeks not only to grow the Linux market, but also to add partners to its network of consultants.

At this stage in the cycle of technology adoption, Linux is at the edge of Geoffrey Moore’s famous Chasm. Currently used by large-firm visionaries who seek a competitive advantage by undertaking internal Linux projects, Linux itself awaits the forces that will carry it across the Chasm from specialty into general acceptance. These forces will be the niche integrators and resellers who will develop Linux-based products for particular specialized business domains, slowly spreading the adoption and use of the new platform. The highly profitable niches, such as oil exploration or financial brokerage, require extensive business domain knowledge. For that matter, so does selling to doctors’ or dentists’ offices. You will be able to track this process of crossing the Chasm by watching for the success stories to shift from triumphs of internal projects to accounts of how niche vendors sell specialized products across multiple companies.

Resellers for vendors who are beginning to put out Linux versions of their software represent a particular challenge. Since the earliest independent software vendors (ISVs) to offer Linux or BSD applications already have UNIX versions, it is likely that the resellers also support UNIX. But in the cases in which a non-UNIX company announces a Linux version, some resellers who lack UNIX/Linux resources may resent the need for additional personnel and expertise, and the burden of supporting an additional platform for the product. In this case they may decide to drop the product, a disincentive that makes software vendors think twice before announcing a Linux product. On the other hand, a reseller who chooses not to support the new Linux versions takes the risk of having the entire business pulled into the vendors’ growing consolidation of the sales channel.

Whatever the complexities of the Open Source environment, no reseller or integrator who lives by service and customization can afford to overlook the chief value of Open Source software for such a business and its customers: What the customer saves on the license fees can be spent on customization and support.

Support, training, and certification

For years, every time someone suggested that businesses should take Linux seriously, someone else was there to say, "But Linux can’t succeed in business until it has credible support organizations." That day is finally here. Enterprises want 24-7 support, and someone to help with problems and to be held accountable.

The original Linux support, asking Linux developers and users via an Internet list how to solve a problem, still provides expert answers, although not necessarily as soon as the question is asked. IT managers do not want to deal with amorphous groups or with many individual outsiders; they want a single point of contact, and they don’t see the point in free support: they want a businesslike business relationship.

When you consider that anyone can enter the support business provided he or she can supply answers to problems via FAX, e-mail, or (most expensive of all) telephone, you may ask why there were not more support businesses earlier. The answer is that large corporations would not have taken such small organizations seriously. It is perfectly possible for a company to be built on an Open Source project, manage that project to the satisfaction of the community, and still see the majority of the support revenue for that project go to other, particularly larger, firms. The first Linux support that was credible to the enterprise came from other large corporations: IBM, Hewlett-Packard, and Compaq, for instance. Nortel Information Network now guarantees 99.5 percent uptime to the thousands of customers of its (Linux-based) Internet services.

Just because there are large players in the support game does not mean there is no place for smaller players. IBM uses its own personnel for first-line support and for consulting services, but relies on Red Hat for next-level support. Dell offers support for the Red Hat Linux it ships on its systems, but the service is provided through LinuxCare (http://www.linuxcare.com), which supports all of the major Linux distributions. The support market is quite fluid. At one time MandrakeSoft contracted with an American consulting company for support of its own boxed Linux-Mandrake distro in the U.S., while Macmillan contracted with LinuxCare to support the Macmillan version of Linux-Mandrake distro. Now that MandrakeSoft is letting Macmillan handle all Linux-Mandrake boxed sales in the U.S., MandrakeSoft no longer contracts for support of its boxed distro, but it does publicize a network of Linux-Mandrake consulting companies that its users might want to try.

LinuxCare itself is a good example of how Linux enthusiasts can found a company and capitalize on their expertise. Any reseller or integrator who wants to keep a long-term relationship with a customer needs to consider starting a support service, and many do, but LinuxCare started with support services, and is branching into consulting and training. The fact that Linux source code is available makes supporting the product easier and more effective. Besides the ability to communicate with customers, a support organization needs a means of tracking bugs and other problems, and organizing the solutions into a knowledge base. In the spirit of Open Source sharing, LinuxCare and Red Hat make their knowledge base available on the Web to all comers. Reciprocally, Linux support businesses can harness the help resources of the entire Linux community in a way that individual businesses would find difficult to manage.

A new support company squarely aimed at the enterprise market is Mission Critical Linux (http://www.missioncriticallinux.com), founded by entrepreneurs with enterprise support experience in DEC UNIX, and dedicated to support of high-availability, high-performance Linux in the enterprise.

Just as VARs and integrators move naturally into support services, support operations can extend themselves into training. The rapid rise of interest in Linux and the shortage of experienced Linux systems administrators have led to another growth industry: Linux training. The focus of modern computer training is not merely competency, but certification. Because businesses set a value on certified personnel — one of the sneers on the Microsoft Linux Myths site asks "how many certified engineers are there for Linux?" — certification operations have also been developing in the Linux community.

There are at least three current projects giving Linux certification tests, and more are on the way. The community-sponsored certification project is the Linux Professional Institute (http://www.lpi.org), funded by several members of Linux International (http://www.li.org). Because it functions as a standards body, the LPI moved more slowly in preparing certification requirements and beginning to administer examinations. Red Hat was the first to offer a certification program; it specializes in that distribution. A more general certification program is already up and running at SAIR (http://www.linuxcertification.com), and Caldera’s Certified Linux Engineer program is also distribution-neutral. As certification grows, the established computer training businesses will also add Open Source platforms and products to their curriculum to train the personnel necessary to a growing support business.

Next chapter: The Applications: Non-Traditional Business Models

Table of Contents

Copyright © 2000 by IDG Books Worldwide, Inc. Published under IDG Books Worldwide, Inc. Open Content License.